I have an MVC application that calls several different web services. While running load tests on the application, I noticed that performance (both throughput and response times) degraded quickly as the concurrent users increased. By process of elimination, I stripped everything out except for a single dummy call to the web service which just returns an empty object and produces a tiny bit of html.
Hitting that test page from another server (using Pylot) with 100 concurrent users, I peak at around 9,000 requests per minute. By comparison, a static html (~70k) peaks at around 20,000 requests per minute. I also created another test page that calls another service (running Solr) and peak at 18,000 requests per minute for a simple query.
I know that it's hard to interpret much from these "results", so I'm looking for any information on how to tune a WCF Web Service for optimum performance under load.